[Qemu-devel] [PATCH RFC] msix_init: input params *_offset isn't the real one

[Qemu-devel] [PATCH RFC] msix_init: input params *_offset isn't the real one

[Cao jin]

The parameter table_offset & pba_offset is kind of confusing, they shouldn't
include bir field.

Signed-off-by: Cao jin <caoj.fnst@cn.fujitsu.com>
---

According to the passed arguments, I guess all the callers of msix_init()
has the same feeling with me, but I am not quite sure about this, so, RFC.

 hw/pci/msix.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/hw/pci/msix.c b/hw/pci/msix.c
index 0ec1cb1..3a16d83 100644
--- a/hw/pci/msix.c
+++ b/hw/pci/msix.c
@@ -264,8 +264,7 @@ int msix_init(struct PCIDevice *dev, unsigned short nentries,
     if ((table_bar_nr == pba_bar_nr &&
          ranges_overlap(table_offset, table_size, pba_offset, pba_size)) ||
         table_offset + table_size > memory_region_size(table_bar) ||
-        pba_offset + pba_size > memory_region_size(pba_bar) ||
-        (table_offset | pba_offset) & PCI_MSIX_FLAGS_BIRMASK) {
+        pba_offset + pba_size > memory_region_size(pba_bar)) {
         return -EINVAL;
     }
 
@@ -282,8 +281,8 @@ int msix_init(struct PCIDevice *dev, unsigned short nentries,
     dev->msix_entries_nr = nentries;
     dev->msix_function_masked = true;
 
-    pci_set_long(config + PCI_MSIX_TABLE, table_offset | table_bar_nr);
-    pci_set_long(config + PCI_MSIX_PBA, pba_offset | pba_bar_nr);
+    pci_set_long(config + PCI_MSIX_TABLE, (table_offset << 3) | table_bar_nr);
+    pci_set_long(config + PCI_MSIX_PBA, (pba_offset << 3) | pba_bar_nr);
 
     /* Make flags bit writable. */
     dev->wmask[cap + MSIX_CONTROL_OFFSET] |= MSIX_ENABLE_MASK |
-- 
2.1.0

Re: [Qemu-devel] [PATCH RFC] msix_init: input params *_offset isn't the real one

[Cao jin]

Sorry for the noise.

On 08/10/2016 11:18 AM, Cao jin wrote:
> The parameter table_offset & pba_offset is kind of confusing, they shouldn't
> include bir field.
>
> Signed-off-by: Cao jin <caoj.fnst@cn.fujitsu.com>
> ---
>
> According to the passed arguments, I guess all the callers of msix_init()
> has the same feeling with me, but I am not quite sure about this, so, RFC.
>
>   hw/pci/msix.c | 7 +++----
>   1 file changed, 3 insertions(+), 4 deletions(-)
>
> diff --git a/hw/pci/msix.c b/hw/pci/msix.c
> index 0ec1cb1..3a16d83 100644
> --- a/hw/pci/msix.c
> +++ b/hw/pci/msix.c
> @@ -264,8 +264,7 @@ int msix_init(struct PCIDevice *dev, unsigned short nentries,
>       if ((table_bar_nr == pba_bar_nr &&
>            ranges_overlap(table_offset, table_size, pba_offset, pba_size)) ||
>           table_offset + table_size > memory_region_size(table_bar) ||
> -        pba_offset + pba_size > memory_region_size(pba_bar) ||
> -        (table_offset | pba_offset) & PCI_MSIX_FLAGS_BIRMASK) {
> +        pba_offset + pba_size > memory_region_size(pba_bar)) {
>           return -EINVAL;
>       }
>
> @@ -282,8 +281,8 @@ int msix_init(struct PCIDevice *dev, unsigned short nentries,
>       dev->msix_entries_nr = nentries;
>       dev->msix_function_masked = true;
>
> -    pci_set_long(config + PCI_MSIX_TABLE, table_offset | table_bar_nr);
> -    pci_set_long(config + PCI_MSIX_PBA, pba_offset | pba_bar_nr);
> +    pci_set_long(config + PCI_MSIX_TABLE, (table_offset << 3) | table_bar_nr);
> +    pci_set_long(config + PCI_MSIX_PBA, (pba_offset << 3) | pba_bar_nr);
>
>       /* Make flags bit writable. */
>       dev->wmask[cap + MSIX_CONTROL_OFFSET] |= MSIX_ENABLE_MASK |
>

-- 
Yours Sincerely,

Cao jin

Re: [Qemu-devel] [PATCH RFC] msix_init: input params *_offset isn't the real one

[Marcel Apfelbaum]

On 08/10/2016 06:18 AM, Cao jin wrote:
> The parameter table_offset & pba_offset is kind of confusing, they shouldn't
> include bir field.
>
> Signed-off-by: Cao jin <caoj.fnst@cn.fujitsu.com>
> ---
>

Hi,

> According to the passed arguments, I guess all the callers of msix_init()
> has the same feeling with me, but I am not quite sure about this, so, RFC.
>
>  hw/pci/msix.c | 7 +++----
>  1 file changed, 3 insertions(+), 4 deletions(-)
>
> diff --git a/hw/pci/msix.c b/hw/pci/msix.c
> index 0ec1cb1..3a16d83 100644
> --- a/hw/pci/msix.c
> +++ b/hw/pci/msix.c
> @@ -264,8 +264,7 @@ int msix_init(struct PCIDevice *dev, unsigned short nentries,
>      if ((table_bar_nr == pba_bar_nr &&
>           ranges_overlap(table_offset, table_size, pba_offset, pba_size)) ||
>          table_offset + table_size > memory_region_size(table_bar) ||
> -        pba_offset + pba_size > memory_region_size(pba_bar) ||
> -        (table_offset | pba_offset) & PCI_MSIX_FLAGS_BIRMASK) {
> +        pba_offset + pba_size > memory_region_size(pba_bar)) {
>          return -EINVAL;

I think we should keep the '(table_offset | pba_offset) & PCI_MSIX_FLAGS_BIRMASK)'
test since it is required by spec, please see: PCI Spec "6.8.2.4. Table Offset/Table BIR for MSI-X"

  Table offset: ...The lower 3 Table BIR bits are masked off (set to zero) by software
                to form a 32-bit QWORD -aligned offset.

This function gets the offset parameters as the whole 32-BIT QWORD and checks it does not collide
with the BIR offset.


>      }
>
> @@ -282,8 +281,8 @@ int msix_init(struct PCIDevice *dev, unsigned short nentries,
>      dev->msix_entries_nr = nentries;
>      dev->msix_function_masked = true;
>
> -    pci_set_long(config + PCI_MSIX_TABLE, table_offset | table_bar_nr);
> -    pci_set_long(config + PCI_MSIX_PBA, pba_offset | pba_bar_nr);
> +    pci_set_long(config + PCI_MSIX_TABLE, (table_offset << 3) | table_bar_nr);
> +    pci_set_long(config + PCI_MSIX_PBA, (pba_offset << 3) | pba_bar_nr);
>

Here is a similar issue. Your interpretation suggests we need to shift left the offset
to make room for BIR, but I think current implementation looks at it differently already
receiving the offset as a 32-bit QWORD and simply does not "look" to the lower bits
implying them 0.

Thanks,
Marcel

>      /* Make flags bit writable. */
>      dev->wmask[cap + MSIX_CONTROL_OFFSET] |= MSIX_ENABLE_MASK |
>

Re: [Qemu-devel] [PATCH RFC] msix_init: input params *_offset isn't the real one

[Cao jin]

On 08/18/2016 06:54 PM, Marcel Apfelbaum wrote:
> On 08/10/2016 06:18 AM, Cao jin wrote:
>> The parameter table_offset & pba_offset is kind of confusing, they
>> shouldn't
>> include bir field.
>>
>> Signed-off-by: Cao jin <caoj.fnst@cn.fujitsu.com>
>> ---
>>
>
> Hi,
>
>> According to the passed arguments, I guess all the callers of msix_init()
>> has the same feeling with me, but I am not quite sure about this, so,
>> RFC.
>>
>>  hw/pci/msix.c | 7 +++----
>>  1 file changed, 3 insertions(+), 4 deletions(-)
>>
>> diff --git a/hw/pci/msix.c b/hw/pci/msix.c
>> index 0ec1cb1..3a16d83 100644
>> --- a/hw/pci/msix.c
>> +++ b/hw/pci/msix.c
>> @@ -264,8 +264,7 @@ int msix_init(struct PCIDevice *dev, unsigned
>> short nentries,
>>      if ((table_bar_nr == pba_bar_nr &&
>>           ranges_overlap(table_offset, table_size, pba_offset,
>> pba_size)) ||
>>          table_offset + table_size > memory_region_size(table_bar) ||
>> -        pba_offset + pba_size > memory_region_size(pba_bar) ||
>> -        (table_offset | pba_offset) & PCI_MSIX_FLAGS_BIRMASK) {
>> +        pba_offset + pba_size > memory_region_size(pba_bar)) {
>>          return -EINVAL;
>
> I think we should keep the '(table_offset | pba_offset) &
> PCI_MSIX_FLAGS_BIRMASK)'
> test since it is required by spec, please see: PCI Spec "6.8.2.4. Table
> Offset/Table BIR for MSI-X"
>
>   Table offset: ...The lower 3 Table BIR bits are masked off (set to
> zero) by software
>                 to form a 32-bit QWORD -aligned offset.
>
> This function gets the offset parameters as the whole 32-BIT QWORD and
> checks it does not collide
> with the BIR offset.
>

Hi Marcel,
     Thanks very much for pointing the accurate reference out.
     I also checked how kernel code handle this field, it is just like 
you said.

     Sorry for this noise.

     Cao jin
>
>>      }
>>
>> @@ -282,8 +281,8 @@ int msix_init(struct PCIDevice *dev, unsigned
>> short nentries,
>>      dev->msix_entries_nr = nentries;
>>      dev->msix_function_masked = true;
>>
>> -    pci_set_long(config + PCI_MSIX_TABLE, table_offset | table_bar_nr);
>> -    pci_set_long(config + PCI_MSIX_PBA, pba_offset | pba_bar_nr);
>> +    pci_set_long(config + PCI_MSIX_TABLE, (table_offset << 3) |
>> table_bar_nr);
>> +    pci_set_long(config + PCI_MSIX_PBA, (pba_offset << 3) | pba_bar_nr);
>>
>
> Here is a similar issue. Your interpretation suggests we need to shift
> left the offset
> to make room for BIR, but I think current implementation looks at it
> differently already
> receiving the offset as a 32-bit QWORD and simply does not "look" to the
> lower bits
> implying them 0.
>
> Thanks,
> Marcel
>